Historical biogeography of the hyperdiverse hidden snout weevils (Coleoptera, Curculionidae, Cryptorhynchinae)

The first dated phylogeny of the weevil subfamily Cryptorhynchinae is presented within a framework of Curculionoidea. The inferred pattern and timing of weevil family relationships are generally congruent with previous studies, but our data are the first to suggest a highly supported sister‐group relationship between Attelabidae and Belidae. Our biogeographical inferences suggest that Cryptorhynchinae s.s. originated in the Late Cretaceous (c. 86 Ma) in South America. Within the ‘Acalles group’ and the ‘Cryptorhynchus group’, several independent dispersal events to the Western Palaearctic via the Nearctic occurred in the Late Cretaceous and Early Paleogene. A second southern route via Antarctica may have facilitated the colonization of Australia in the Late Cretaceous (c. 82 Ma), where a diverse Indo‐Australian clade probably emerged c. 73 Ma. In the Early Eocene (c. 50–55 Ma), several clades independently dispersed from Australia to proto‐New Guinea, i.e. the tribe Arachnopodini s.l., the ‘Rhynchodes group’ and the genus Trigonopterus. New Zealand was first colonized in the Late Palaeocene (c. 60 Ma). Divergence time estimations and biogeographical reconstructions indicate that the colonization of New Guinea is older than expected from current geological reconstructions of the region.


Introduction
With c. 400 000 described species, beetles are the most species-rich group of known animals. Understanding the mechanisms that govern the assembly of such an astonishing diversity is therefore of great significance. Yet the evolution of many major beetle groups remains little explored due to a lack of fossil-based dated phylogenies. For the economically important and evolutionarily interesting weevils (Curculionoidea), only few studies have attempted to provide sound temporal estimations of divergence times at higher taxonomic ranks (e.g. McKenna et al., 2009;Gunter et al., 2016;Shin et al., 2018). Delimitation of many larger weevil subfamilies and tribes is often ambiguous, and current classifications are mainly based on ad hoc decisions rather than phylogenetic reconstructions Correspondence: Alexander Riedel, Museum of Natural History Karlsruhe, D-76133 Erbprinzenstr. 13, Karlsruhe, Germany. E-mail: riedel@smnk.de (Oberprieler et al., 2007. This often hampers the compilation of reliable datasets to estimate divergence times and, consequently, comprehensive phylogenetic reconstructions and divergence dating are restricted to few well-defined weevil subfamilies, such as Platypodinae (Jordal et al., 2011;Jordal, 2015), Apioninae (Winter et al., 2017), and Ceutorhynchinae (Letsch et al., 2018). Another challenge for weevil dating is the choice of reliable fossil calibrations. Weevil fossils are legion, but many of these cannot be assigned to extant weevil families or subfamilies without contention. Legalov (2012) compiled an overview of weevil fossils from the Mesozoic, with several recent updates (Legalov, 2014a(Legalov, , 2014b. However, the assignment of many of these fossils to extant families is still questionable and under debate (e.g. Oberprieler et al., 2014;Gunter et al., 2016), leaving only a handful of suitable fossils to use in divergence time dating analyses.
Cryptorhynchinae (hidden snout weevils) are one of the most diverse groups of Curculionidae, themselves one of the two most species-rich families on Earth (Grebennikov & Newton, 2009). They share a unique morphological feature that distinguishes them from most other weevil groups. As a defensive pose, they retract their rostrum into a canal formed by proand mesosternal structures while their legs are usually folded in a ventral position. Thus, feigning death, they often simulate natural objects, such as seeds, small stones or twigs (Lyal, 1993;van de Kamp et al., 2014).
The most comprehensive approach to address the phylogenetic relationships of Cryptorhynchinae included 105 ingroup genera representing all geographic regions . This study tested the monophyly of Cryptorhynchinae and revealed a monophyletic Cryptorhynchinae s.s., excluding the tribes Aedemonini and Camptorhinini. The taxonomic status and classification of Cryptorhynchinae are under discussion because their main character of morphological identification, i.e. a rostral furrow combined with a mesosternal receptacle, is prone to convergence (Lyal, 2014;. The choice of uniting them within the polyphyletic subfamily 'Molytinae' (Oberprieler et al., 2007;Lyal, 2014) hardly improved the situation.
Taxonomic diversity of Cryptorhynchinae peaks in the Australian and Neotropical regions, followed by the Pacific Islands, and then the Oriental and the Holarctic regions. Cryptorhynchinae (s.s.) appear largely absent from the Afrotropics, where they seem to be replaced by the tribe Aedemonini (Molytinae). Even small isolated islands may host substantial radiations (Paulay, 1985). Based on the high percentage of new species added by recent taxonomic revisions, a total of > 15 000 Cryptorhynchinae species can be anticipated (e.g. Eberle et al., 2012;Setliff, 2012;Tänzler et al., 2012;Riedel et al., 2013Riedel et al., , 2014Luna-Cozar et al., 2014;Riedel & Narakusumo, 2019). Recent studies on the Western Palaearctic Cryptorhynchinae of the Acalles group (Astrin & Stüben, 2008;Astrin et al., 2012) and the Indo-Australian genus Trigonopterus Fauvel Toussaint et al., 2017b) provided insights into their evolution, but the systematics and evolution of the highly diverse South American and Indo-Australian faunas remain largely unexplored. Many species and genera of the litter fauna are still undescribed, while the relationships and composition of major groups are in equal need of study.
The current classification of Cryptorhynchinae s.s. is more than problematic: as most of the established tribes and subtribes, such as Gasterocercini, Tylodina and Mecistostylina, appear to be polyphyletic,  advocated for the use of Cryptorhynchinae s.s. without any subcategories. Some biogeographically defined groups appeared highly supported, i.e. a large Indo-Australian clade or a smaller clade comprising the majority of the New Zealand fauna, but these cannot be named formally unless a larger portion of the existing genera can be assigned and/or characters are identified that allow their morphological diagnosis.
Estimates of reliable divergence times of major groups of Cryptorhynchinae are still missing. However, methods inferring the potentially differential diversification among clades, i.e. speciation and extinction over space and time, or the impact of specific traits (e.g. lifestyle features, morphological characters or geographical distributions) as driving forces on diversification, rely on the analyses of dated phylogenetic trees sufficiently representing the species richness of focal clades (e.g. Morlon, 2014;Ng & Smith, 2014;Maddison & FitzJohn, 2015;Rabosky & Goldberg, 2015). Studies such as the ones focusing on the evolutionary history of the extremely diverse Trigonopterus, possibly with > 1000 species in New Guinea alone, also depend on sound estimates of their evolutionary age. Thus, the retrieval of a robust maximum age for Trigonopterus is one goal of the present study. As the sister group of Trigonopterus remains unknown but is presumably found among the wingless genera of Cryptorhynchini, i.e. 'Tylodina', we tried to include as many lineages of them as possible. A large portion of these edaphic species is still undescribed, even at genus level, which leads to an unusually high number of unidentified taxa contained in the dataset. In some cases, taxonomic problems preclude a robust identification (Riedel, 2017). Arachnobas Boisduval is a peculiar genus recently recognized as belonging to the Indo-Australian clade of Cryptorhynchinae . It is endemic to the Papuan region and absent from Australia, and thus a likely candidate of a radiation confined to New Guinea or a Proto-New Guinea insular setting. As such, it may have a similar history of diversification as Trigonopterus and, in combination, both taxa may provide insights into the biogeographic history of this area.
The goals of the present study are to present a robust phylogeny of Cryptorhynchinae with comprehensive taxon sampling of Cryptorhynchinae s.s. from all major geographic regions (this forms the basis for a revised classification) and to generate reliable divergence time estimates and historical biogeography of major clades within the group.

Phylogenetic analyses
Alignment procedures for all protein-coding and ribosomal RNA genes were separately conducted with the online version of the program mafft v.7.409 (Katoh & Standley, 2013;Katoh et al., 2017), applying the automatic method search (protein-coding genes, FFT-NS-1 method; rRNA genes, L-INS-i method). Alignments of ribosomal RNA genes are challenging, as positional homology of variable regions is hard to obtain. We therefore excluded ambiguous positions in all ribosomal RNA alignments with the software aliscore v.2.0 (Misof & Misof, 2009). The alignments of all genes were subsequently assembled using the software FASCONCAT v.1.0 (Kück & Meusemann, 2010). Codon positions of each protein-coding gene, as well as each ribosomal RNA gene were defined as distinct partitions a priori. This resulted in a dataset comprising 5690 nucleotides and 18 partitions.
We used modelfinder as implemented in iq-tree v.1.6.10 (Nguyen et al., 2015;Chernomor et al., 2016;Kalyaanamoorthy et al., 2017) to find the best-fitting partitioning and model scheme. Due to small partitions, we deliberately refrained from using the free-rate model approach in iq-tree (B. Q. Minh, personal communication), and also restricted the model search solely to those models supported by the Bayesian inference (BI) software package beast  for both maximum likelihood (ML) and BI analyses. For ML tree reconstruction analyses, we used iq-tree v.1.6.10. Based on the detected partition-model scheme, we performed 100 independent tree searches with a random start tree and decreased perturbation strength (−pers 0.2). All analyses were run with edge-proportional partition models (−spp). Nodal support was assessed using 1000 ultrafast bootstrap replicates (UFBoot; Minh et al., 2013), with the 'bnni' option to reduce the risk of overestimating branch support (Hoang et al., 2018), and an increased maximum number of iterations to stop (−nm 10 000). Additionally, we also performed 1000 replicates of the Shimodaira-Hasegawa-like approximate likelihood ratio test (SH-aLRT, Guindon et al., 2010).

Divergence time estimation
Divergence times were estimated in a Bayesian Markov chain Monte Carlo (MCMC) framework, using the software beast v.1. 10.2 (Lemey et al., 2018). For all beast analyses we used the topology from the best ML tree obtained by iq-tree as starting tree and constrained the monophyly of all families, except Anthribidae, which was polyphyletic in the ML analyses. Instead we constrained the monophyly of Urodontinae and Anthribinae. Each analysis was run for 125 million generations (sampling every 10 000 generations). The number of generations discarded as burn-in was based on the examination of posterior distributions in tracer v.1.7.1 (Rambaut et al., 2014). Post burn-in samples were combined across runs to summarize parameter estimates and used to generate a maximum clade credibility (MCC) tree with median node heights using treeannotator v.1. 10.2 (Lemey et al., 2018).
To test the impact of different tree priors, clock model partitioning, fossil calibration schemes, and fossil calibration prior densities on the age estimations of Cryptorhynchinae, we conducted eight independent MCMC analyses (Table 1). In a first setup (C0), we compared different tree models, i.e. diversification process priors, using either a Yule (pure-birth) tree prior (C01, C05) or a birth-death model (BD) prior (C02, C06). The partitioning scheme and models of nucleotide substitution were the same as for the ML analyses. For the clock model priors, we used the uncorrelated lognormal relaxed-clock (UCLN) model (Drummond et al., 2006). In the different analyses, the clock models were either linked (C01, C02) or unlinked (C05, C06) among the partitions. To test the fit of different parameter settings, we used Bayes factors (BFs), obtained by marginal likelihood estimations (MLEs) of all four analyses, using the path sampling (PS) and stepping-stone sampling (SS) methods in beast with default parameter settings (Baele et al., 2012). Using the resulting best model scheme, we ran additional analyses with the fossil calibration schemes described in the following.
To calibrate the relaxed clocks in beast, we followed the calibration schemes used in Shin et al. (2018). As our ML   Kuschel (1959) tree reconstruction results differ slightly from those of Shin et al. (2018), we only applied compatible fossil calibrations ( Table 2). Similar to Shin et al. (2018), we tested the effect of two alternative fossils for Entiminae, using three different fossil-calibrating schemes: C0, no Entiminae fossil; C1, including the supposed oldest Entiminae fossil of the genus Dorotheus (Kuschel, 1959); C2, including a younger Entiminae fossil of the genus Polydrusus (Yunakov & Kirejtshuk, 2011). To consider a potential impact of the fossil calibration prior densities on the divergence dating analyses, we independently applied exponential and uniform calibration priors (Ho & Phillips, 2009). Uniform prior estimates were applied with a hard lower bound provided by the minimum age of particular fossil layer intervals ( Table 2). The hard upper bound for the maximum age of Curculionoidea was provided by the age of oldest polyphagan beetle †Leehermania prorova (223 Ma; Chatzimanolis et al., 2012). For the maximum age of Curculionidae + Brentidae, Brentidae and Entiminae, the upper bound was provided by the proposed maximum age of Curculionidae (151 Ma; Oberprieler et al., 2014). Exponential prior estimates were applied with identical hard lower bounds defined by fossil layer intervals and an adapted soft upper bound, so that 95% of the distribution lay between the fossil age and 223 Ma.

Biogeographical analyses
Biogeographical analyses were conducted using biogeobears v.1.1.2 (Matzke, 2013) as implemented in the r v.3.5.3 statistical software (R Development Core Team, 2019). biogeobears estimates ancestral ranges under different models; it uses the dispersal extinction cladogenesis (DEC) model (Ree & Smith, 2008), as well as likelihood interpretations of the dispersal-vicariance analysis (DIVA) model (Ronquist, 1997) and the BAYAREA model (Landis et al., 2013). It further implements a parameter describing founder-event speciation (+J), which allows cladogenetic events where one daughter lineage colonizes a new range via founder-event speciation, while the other retains the ancestral range. While this parameter has been shown to result in higher likelihood compared with models ignoring this parameter (Matzke, 2012(Matzke, , 2014, its use has recently been criticized (Ree & Sanmartín, 2018). Models incorporating +J have the tendency to underestimate anagenetic dispersal events at ancestral nodes in favour of 'jump dispersal', which can potentially distort the ancestral range reconstruction of ancient groups with a proposed widespread distribution, such as Cryptorhynchinae s.s., which are almost cosmopolitan. As the statistical comparison to models excluding founder-event speciation has also been suggested to be inaccurate, we refrained from implementing models including founder-event speciation in the present study. The Akaike information criterion corrected for small sample size was used to compare the fit of all models with the given data (Table 3). Ancestral range reconstructions were estimated using the MCC tree from the best BEAST analysis (see later). Prior to the analysis, all outgroups except Piazurus were removed to avoid an impact of more distant outgroups on the area reconstruction. The number of maximum areas per ancestral range was constrained to three. Studies focusing on ancestral area reconstruction methodology have shown that a larger maximum number of areas led to an overestimation of ancestral area sizes, neglecting the often limited vagility of the studied groups (Kodandaramaiah, 2009(Kodandaramaiah, , 2010. Therefore, we selected the following seven regions for the biogeobears analyses: (A) Palaearctic, (B) Nearctic, (C) Neotropical, (D) Oriental, (E) Australia, (F) (Proto-) New Guinea including Samoa, and (G) New Zealand and New Caledonia. We also generated three time slices to reflect tectonics throughout the Cenozoic following recent palaeogeographic works (Ezcurra & Agnolín, 2012;Seton et al., 2012). Appendix S3 provides details on dispersal probabilities and area connections over time.

Phylogenetic analyses
Results of the MLE runs of the eight different BI analyses are shown in Table 1. Based on BF comparisons between the analyses with two fossil calibrations (C01, C02, C05 and C06), unlinked clock models represented a better fit (C05 and C06). Among the latter, the MLE comparisons were equivocal, as PS and SS sampling methods indicate different best model schemes. SS suggested the analysis with a Yule tree model (C05) as the best (BF = 4.13), whereas PS favoured a BD tree model (C06; BF = 14.85). However, as the effective sample size of log likelihood and other parameters of the C05 analysis did not converge after 125 million generations, we relied on BD tree models in all further analyses with additional fossil calibration points (C11, C13, C21 and C21). However, the analysis with two fossils (C06; shown in Fig. 1) generally shows a better marginal likelihood value than the models with additional calibration points. In the following, we discuss the results of all eight BI runs, as well as the best ML tree reconstruction results.
Nodal supports from UFBoot and SH-aLRT of the best ML analysis, as well as posterior probability (PP) values of the best BI analysis (C06) are provided in the text for the discussed relationships. All tree reconstruction analyses results are provided in Appendix S2.
Both BI and ML tree reconstructions show some differences in higher-level weevil relationships, mainly due to the inconsistent position of Car (Caridae) and Urodontinae, whose relationship generally lack strong nodal support. In the ML analyses, as well as the BI analyses based on linked clock models (C01, C02) and one BI analysis based on unlinked clock models (C23), Anthribidae appeared polyphyletic as Anthribinae are recovered sister to Nemonychidae (ML: SH-aLRT = 94.4, UFBoot = 50), and Urodontinae are recovered sister to a clade comprising the remaining families (ML: SH-aLRT = 96.7, UFBoot = 43), except for Car as the single representative of the family Caridae, which appeared as the first branch in the tree (ML: SH-aLRT = 8.7, UFBoot = 61). By contrast, most analyses based on unlinked clock and BD tree models (C05, C06, C11, C13, C21) recovered Anthribidae (Anthribinae + Urodontinae) as monophyletic (C06, BI: PP = 0.99) and Caridae as sister to the clade Brentidae + Curculionidae (C06, BI: PP = 0.76). The position of Caridae as sister group to the clade Brentidae + Curculionidae in the BI analyses with unlinked clock models, supports its recognition as a distinct family and is consistent with phylogenetic studies based on adult and larval characters (Morrone & Marvaldi, 2000;Marvaldi et al., 2002), as well as recent large-scale molecular analyses (McKenna et al., 2009;Shin et al., 2018). The inconsistent position of Urodontinae among the analyses generally reflects the uncertainty of their phylogenetic placement. The placement of Urodontinae as sister to Anthribinae in most BI analyses corroborates their inclusion into Anthribidae, as proposed by Kuschel (1995) and further recovered by the phylogenomic study of Shin et al. (2018), as well as by a molecular analysis of Australian weevils (Gunter et al., 2016). By contrast, the isolated position of Urodontinae in the ML analyses and their relationship to Attelabidae, or Attelabidae + Belidae in the remaining BI analyses, corroborate Crowson (1984) and Thompson (1992), which placed Urodontidae as a family separate from Anthribidae.
Among the different analyses performed in biogeobears, the DEC model was significantly preferred over the DIVALIKE and BAYAREA models (Akaike weight = 1; shown in Fig. 2). The ancestral range estimated by the DEC model for Cryptorhynchinae s.s. was South America (C = 0.62, CE = 0.16, CF = 0.21). The initial radiation within Cryptorhynchinae s.s. was characterized by a further diversification of the 'Cryptorhynchus group' and its relatives in South America (C = 0.99). Within the 'Cryptorhynchus group' several species independently colonized North America, Eurasia and the Australian region in the Eocene. A similar pattern was recovered for members of the originally Neotropical 'Acalles group' and its relatives (C = 0.86), which also colonized the Western Holarctic even earlier in the Late Cretaceous. A colonization of the Palaearctic from South America was also found, for instance, by Toussaint et al. (2017a) for Hydrophilus water scavenger beetles and can be explained by either long-distance dispersal or range expansion via the Nearctic followed by regional extinction. The occurrence of North American representatives in both groups supports the latter scenario. For the subsequent radiation of Cryptorhynchinae s.s., a range expansion to Australia and Proto-New Guinea was estimated (clade A: C = 0.28, CE = 0.31, CF = 0.40; clade B: CE = 0.41, CF = 0.53) between 73 and 91 Ma, and the origin of the Indo-Australian clade was recovered in Australia and/or Proto-New Guinea (clade C: E = 0.26, F = 0.38, EF = 0.35) at c. 73 Ma, indicating a continental range expansion via dispersal from South America possibly through Antarctica in the Late Cretaceous. This scenario is concordant with a proposed connection between South America and Australia via a land bridge through Antarctica until c. 60 Ma (Scotese, 2004;Seton et al., 2012). This pattern has recently been suggested for several beetle clades using a combination of Bayesian relaxed-clock dating and parametric historical biogeography. For instance, Kim & Farrell (2015) proposed a hypothesis in which Chiasognathini stag beetles expanded their range towards Antarctica in the Cretaceous. Gustafson & Miller (2017) suggested the colonization of Antarctica by Macrogyrus whirligig beetles in the Palaeocene. A similar pattern was suggested for Platynectes diving beetles in the Eocene (Toussaint et al., 2017b), and for Hydrobiusini and Oocyclus water scavenger beetles in the Cretaceous (Toussaint & Short, 2017. This pattern therefore seems to be much more common than previously thought and is supported by recent palaeoclimatic evidence. Antarctica had a much warmer climate during most the Cenozoic due to its connection with other components of the Gondwana supercontinent. As a result, Cenozoic favourable landscapes existed in Antarctica with dense forests (subtropical at times) that could have hosted a diverse fauna before the setup of a polar climate on this land mass (Poole & Cantrill, 2006;Francis et al., 2008). Glaciations only initiated after Australia started rifting away in the Oligocene and triggered ecosystem turnover in Antarctica (Galeotti et al., 2016;McKay et al., 2016). With Australia's position between Antarctica and Proto-New Guinea, a colonization of Australia prior to Proto-New Guinea is plausible. The subsequent early radiation of the Indo-Australian clade in Australia corroborates this hypothesis. However, the occurrence of one Chilean species deeply nested in the 'Indo-Australian clade' indicates that this clade may in fact have evolved in a more widespread Gondwanan range, including South America, possibly in the southern temperate environment of Nothofagus forests. An equally plausible explanation could be a recolonization of southern South America: the case of Strongylopterus distributed in both New Zealand and Chile underlines the potential of dispersal of wood-inhabiting weevils in the subantarctic region, possibly by sea currents. A denser taxon sampling in southern Australia, New Zealand and Chile should be attempted in the future.
Within the 'Indo-Australian clade', subsequent dispersal events to Proto-New Guinea took place three times independently at around the same time, i.e. c. 50-55 Ma, by Arachnopodini s.l., the crown group of Trigonopterus (excluding the T. squamosus group), and the 'Rhynchodes group'. This timing is much earlier than expected and contrasts with geological reconstructions that anticipate the first major land areas not to have emerged before 35 Ma ('peninsular orogeny ';Ufford & Cloos, 2005) or 20 Ma (formation of the northern arc of New Guinea; Hall, 2009), although the first volcanic arcs in the area appeared as early as 60 Ma (Hall, 2009) and the Papuan Ultramafic Belt ophiolite has an age of c. 58 Ma (Baldwin et al., 2012). These latter dates are in line with our current reconstruction and indicate that New Guinea may have acted as a museum of diversity in addition to being a cradle as suggested by recent evolutionary studies focusing on the island fauna (e.g. Unmack et al., 2013;Georges et al., 2014;Toussaint et al., 2014;Janda et al., 2016;Oliver et al., 2017;Lam et al., 2018;Tallowin et al., 2018). Our study brings more evidence to the potential role of New Guinea as an older land mass that may have hosted the early stages of several island clades. For instance, a time-calibrated phylogeny of netwing beetles endemic to New Guinea (Bocek & Bocak, 2019) recovers a similar age (51 Ma). The origin of corvoid birds from New Guinea is dated from the Eocene c. 45 Ma (Jønsson et al., 2011;Aggerbeck et al., 2014). New Guinean endemic mayflies also possibly have originated as early as the Eocene on the island (Cozzarolo et al., 2019). These results suggest that substantial areas may have been subaerial in Proto-New Guinea much earlier than hitherto expected. The age of the Palaeocene 'New Zealand clade' conflicts with the hypothesis of Oligocene marine transgression of New Zealand  Fig. 2. Estimation of the historical biogeography for Cryptorhynchinae s.s. using a dispersal-extinction-cladogenesis model in biogeobears. The coloured boxes represent the seven areas implemented in the palaeogeographical model, as well as the six most important ranges discussed in the text. Pie charts at the nodes of the tree represent the relative probabilities of the ancestral areas. The map represents the historical southern dispersal route from South America to Australia, New Guinea and New Zealand. some 25-23 Ma (Waters & Craw, 2006), which is in line with the multitaxon analysis of Wallis & Jorge (2018).

Conclusion
We reconstructed the biogeographical history of Cryptorhynchinae, with an origin in the Neotropical region during the Cretaceous. Two distinct colonization routes are proposed: a northern route, which led to at least two independent dispersals to both North America and Eurasia, and a southern route, which possibly facilitated the colonization of Australia, New Guinea and New Zealand via Antarctica in the Late Cretaceous. Within the Indo-Australian clade, the reconstructed lengths and divergence times of the early branches are conspicuously short, and many nodes are only moderately supported, leading to incongruent relationship hypotheses between the distinct analyses. This pattern further indicates a rapid radiation of the 'Indo-Australian clade' after its arrival in Australia. Cryptorhynchinae constitute c. 30% of the Australian weevil fauna (Pullen et al., 2014) and further comprise the majority of Australian weevils using dead wood as a food resource. This may indicate that the stage was set for their rapid radiation once they reached the Australian continent. However, 'ancient rapid radiations' phenomena have been proposed to substantially impede phylogenetic reconstructions (Whitfield & Lockhart, 2007;Whitfield & Kjer, 2008). Together with the still highly incomplete taxon sampling of the Indo-Australian fauna (Riedel et al., 2013;Pullen et al., 2014;, scenarios about the evolution of ecological and/or morphological traits, which might have facilitated their radiation, remain uncertain (Franz & Engel, 2010;Gunter et al., 2016). We therefore propose to focus on lower taxonomical levels that allow a denser taxon sampling and thus more precise inferences of diversification pattern. Previous studies on the evolution of the Indo-Australian genus Trigonopterus could already reconstruct several radiations of these weevils in the geologically complex Indo-Australian archipelago. They generally place the colonization of New Guinea, Indonesia and New Caledonia in the Late Miocene Toussaint et al., 2017c. However, these studies did not infer the divergence times of Trigonopterus in a taxonomically larger context and could not therefore implement calibration fossils. The proposed age of Trigonopterus (c. 54 Ma) recovered in the present study, however, indicates a much older diversification of this genus. With this age estimation at hand, and in combination with the ongoing taxonomic research (Riedel et al., 2013Riedel & Narakusumo, 2019), future studies on the evolution of the genus Trigonopterus could help to elucidate the Cenozoic history of Cryptorhynchinae weevil diversification in the Indo-Australian regions.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Appendix S1. List of specimens, markers, and GenBank accession numbers.
Appendix S2. Details on the dispersal rate scaler and adjacency matrices applied in biogeobears.
Appendix S3. Results of the eight Bayesian Inference (BI) runs in BEAST, as well as the best tree of 100 IQ-TREE analyses.